FIG. 1 shows a telephone network constituting a star-net-topology. A fibre backbone is terminated at a Central Office (CC)) supporting 500-20000 end customers. From the CO primary cables carrying 100-1200 twisted pairs runs to Cabinets (Cab) which are cross-connect-points that normally have no power and environmental capabilities. The last 100-800 meters of twisted pairs between the Cab and the customer premises is called the distribution network.
It is desirable to re-use the existing copper network for delivering high capacity data connections, i.e. broadband access, to the customer premises. The family of systems designed for this purpose is called Digital Subscriber Lines (DSL) systems. Example DSL technologies (sometimes called xDSL) include High Data Rate Digital Subscriber Line (HDSL), Asymmetric Digital Subscriber Line (ADSL), a version of DSL with high download speed than upload speed, Very-high-bit-rate Digital Subscriber Line (VDSL).
The most recently standardized broadband technology for the copper networks is VDSL. It provides higher data bandwidths than precursors like ADSL and HDSL, but to the expense of shorter reach. Currently, the latest version of the VDSL-standard is called VDSL2.
For VDSL, network operators can only partly use the same deployment strategy as for ADSL, which is to install the DSL-modems in the central office. From the central office, VDSL can be offered to 30-50% of customers compared to 80-90% in the case of ADSL, dependent on the topology of the specific network. To further increase the VDSL customer base, it can be deployed from a fibre-to-the-cabinet (FTTCab) infrastructure meaning that the fibre termination point is moved closer to the premises giving a shorter copper loop. The cabinet is deployed at the local cross connect point for the distribution network, which normally is the only point-of-presence for the cable. The VDSL, digital subscriber line access multiplexer (DSLAM) equipment, where all DSL-moderns are connected to the backbone network, will be placed in the new cabinet and VDSL is used to serve the customers over the last drops of cable.
Statistics of cable lengths and the network topology are crucial parameters when deploying DSL. The copper loops have the property that the possible data capacity that can be transferred decreases for longer loops. A second property that limits the possible data rates is crosstalk, i.e. self made noise that occurs between different loops in the same cable during transmission. This effect is more pronounced on shorter loops, since one important kind of crosstalk tends to decrease with increasing loop length.
Common to all existing DSL systems is that they are designed for a worst case scenario. This means that the systems are designed for a maximum cross-talk scenario, i.e., that all systems are transmitting all the time and that they generate full cross talk to each other.
There are two kinds of crosstalk: Near End Cross Talk (NEXT) and Far End Cross Talk (FEXT). The NEXT is noise that comes from a transmitter on a neighbouring pair at the same end of the line and the FEXT is noise that comes from a transmitter on a neighbouring pair located at the far end of the line.
The NEXT is always stronger than the FEXT and most DSL systems are designed to avoid the NEXT but assumes that there are always FEXT present. Some systems operating at low frequencies (e.g. less than 500 kHz) are designed to take into account also NEXT. This is possible since NEXT is not very severe at low frequencies, which is illustrated in FIG. 3.
By coordinating the signal transmission and reception for a plurality of modems in the CO, the FEXT can be eliminated. This is often referred to as vectored transmission, vectoring, or vectored DSL. For shorter loops the FEXT is the dominating noise source of essentially the entire frequency band. Thus, the elimination of FEXT, with vectored DSL, can substantially increase the achievable bitrates, especially for modem on shorter loops. Deployment, of VDSL from the FTTCab as discussed above will lead to much shorter loops and also fewer loops in each cable. With vectored VDSL the bitrates for shorter loops (<800 m) can be increased with 50% to 200% depending on loop length. Shorter loops have normally higher levels of FEXT than longer loops, and can therefore gain more when removing the FEXT with vectoring techniques.
Coordinated signal transmission and reception of all modems, referred to as vectored DSL, is possible, since the modems are co-located in a CO or cabinet. In the upstream direction (signal reception) this is called FEAT cancellation or multi-user detection. In the downstream direction (signal transmission) it is called FEXT pre-coding, but sometimes it is also called FEXT cancellation in the downstream direction.
There exist numerous techniques for vectored transmission (i.e. the use of FEXT pre-coding and multi-user detection). However, since the FEAT vector channel has property called row-wise diagonal dominance, it has been proved that diagonalizing pre-coding for the downstream and zero forcing equalization for the upstream yields close to optimal performance which is further described in R. Cendrillon, M. Moonen, E. Van den Bogaert, G. Ginis, “The Linear Zero-Forcing Crosstalk Canceller is Near-optimal in DSL Channels”, in Proc. of IEEE Global Comm. Conf. (GLOBECOMM), Dallas, Tex., pp 2334-2338, November 2004.
The drawback with all vectoring techniques is that they lead to highly complex and large chipsets and systems. The complexity of the vectoring processing grows with the square of the number of modems the vectored DSL system can handle.
To guarantee best possible bit-rate performance all modems in a cable must be part of the same vectored DSL system. If some non-vectored DSL-modems are operating on loops in the same cable as the vectored DSL-modems they will generate FEXT, which can reduce the bit rates significantly for some or all of the vectored DSL-modems.
Since it can be practically impossible to build a vectored DSL system that can handle a large number of DSL-modems suboptimal solutions must be used. A simple solution is to not allow more DSL-modems to communicate in a cable than what the vectored DSL-system can handle.
Other solutions are based on the fact that most often, any single DSL-modem receive strong FEXT from only a few other loops DSL-modems). But, which loops that generates strong crosstalk depends both on frequency and which loop that is considered. An electronically controlled cross-connecting device can be used to select a subset of loops which receive as little crosstalk as possible from other loops.
Thus, a problem with the existing solutions is either that the number of DSL modems is limited to the capacity of the DSL vector system (i.e. the pre-coder and the detector), or if the number of DSL modems is not limited by the capacity of the DSL vector system, some of the DSL modems will not be affected by the DSL vector system (denoted non-vectored DSL modems) which results in that they will introduce cross-talk and some of the DSL-modems that are operated by the DSL vector system cannot benefit from the vectoring since they will receive crosstalk from the non-vectored DSL modems.
Another problem with the latter case of having a mixture of non-vectored DSL-modems and vectored DSL modems is that unless all FEXT-couplings between all wires in the cable is more or less exactly known, it will be very difficult to predict what bitrates a certain customer using a certain modem can achieve and consequently what services an operator can offer to that customer.